IA-ICGCN: Integrating Prior Knowledge via Intra-event Association and Inter-event Causality for Chinese Causal Event Extraction

Abstract

AbstractCausal event extraction (CEE) is a joint extraction task of event detection and event causality discrimination, which is of great value on dialogue system, event prediction and so on. However, the ambiguity of Chinese event description and the implicit causal relationship makes the poor performance with existing models. To deal with this problem, we propose a model to incorporate domain knowledge by taking Intra-event Association and Inter-event Causality into account. We use causal indicators to obtain sentences containing causal relationships from domain texts, and use these sentences to construct event association networks and causality transition networks. To obtain complete event expressions and accurate event causality, we use graph convolutional neural networks (GCN) to encode the information of the two networks separately. Finally, all the information are fed into the Bidirectional Long Short-Term Memory networks. Experimental results on two datasets show that our method outperforms the state-of-the-art baselines.KeywordsEvent detectionCausal event extractionChinese eventGraph convolutional neural network